The German Impersonal Passive from the Perspective of Construction Grammar ()
1. Introduction
In German, there is a special passive construction known as the impersonal passive, in which the verb is intransitive and the agent is omitted. Concrete examples are provided below:
(1) Hier wird gearbeitet. (Work is being done here.)
(2) Da wurde laut gequietscht. (There was loud squealing there.)
(3) Es wurde geplaudert u. photographiert. (There was chatting and photographing.)
(4) Aber geblüht wird nur, wenn die Pflanze auch etwas älter ist. (But it only blooms when the plant is a bit older.)
It is worth noting that the verb blühen (to bloom) is unaccusative and only weakly agentive, yet it can still be passivized in this context.
Traditional grammatical descriptions of the German impersonal passive are both incomplete and unclear in distinguishing between the impersonal passive and the personal passive (Curme, 1905; Behaghel, 1924; Paul, 1958, 1959; Brinkmann, 1962; Drosdowski, 1984; Helbig & Buscha, 1988). In traditional grammars, the impersonal passive is treated in a fragmentary and cursory manner (Curme, 1905; Behaghel, 1924; Paul, 1959; Brinkmann, 1962). This theoretical gap served as the point of departure for subsequent research.
The promotion analysis proposed by Relational Grammar significantly advanced the discussion. However, Keenan (1976) and Keenan and Comrie (1977) criticized this promotional approach with the demotion analysis, with Keenan and Comrie (1977) being the most influential. He pointed out that demotion can still occur even when no object is promoted to subject position. Using examples from Finnish, he demonstrated that there indeed exist passives without promotion, that is, subjectless passives. In such cases, an intransitive verb appears and consequently no promotion to subject takes place. Nevertheless, this analysis remains purely syntactic and does not address the licensing conditions of intransitive verbs in the impersonal passive (Keenan & Comrie, 1977).
Subsequently, researchers began to focus on semantic features. Moorcroft (1985: p. 160) proposed that verbs capable of forming impersonal passives share the semantic feature of a “highly volitional agent”. The agent need not appear in the impersonal sentence; it is sufficient that volitionality is present in the verb’s lexical entry. It is precisely this restriction to volitional agents that results in passives retaining semantic information about the active subject (Moorcroft, 1985). Moorcroft’s appeal to the lexicon is a call for semantic factors to determine which verbs allow impersonal passives, and therefore her analysis is not incompatible with Relational Grammar. In fact, she aligns with other scholars who have also proposed volitionality as an important feature (cf. Kirsner, 1979; Perlmutter, 1978; Fagan, 1992; Shannon, 1992).
Arnett (2002: p. 158) draws on Hopper and Thompson’s (1980) Transitivity Hypothesis to offer an account of the licensing conditions for intransitive verbs in the impersonal passive. According to her analysis, the impersonal passive involves only one participant. Nevertheless, it exhibits features associated with transitivity, with the volitionality of that single participant serving as a relevant factor. She further points out that the impersonal passive and the personal passive differ along two basic dimensions of transitivity: the number of participants and punctuality. A prototypical passive sentence encodes a transitive event involving two or more distinct participants. In Arnett’s view, the impersonal passive involves a setting selected as a trajector and an activity that takes place within that setting. In the personal passive, the focus shifts away from the agent and onto a participant located downstream in the flow of energy; this participant is then selected as the trajector of the sentence. By contrast, in the impersonal passive, there is only one participant: the trajector of the base process is that single participant, and the process unfolds over a certain period of time, with all parts of the process being profiled. A process that connects a thing and a relation is intransitive (Langacker, 1991: p. 345).
By contrast, Beatrice Primus (2011) offered a different explanation for the passivization of unergative verbs in the impersonal passive. She pointed out that unergative verbs are passivizable neither because their original telic semantics are suppressed by constructional constraints (thereby highlighting atelicity and habituality) nor because of any inherent ambiguity in the verbs themselves. The real reason lies in the fact that agent demotion suppresses the quantificational reading of the arguments, which in turn triggers atelic and habitual interpretations (Primus, 2011).
However, agent demotion is not a syntactic operation exclusive to the impersonal passive. The personal passive also requires it, and there is no necessary connection between demoting the agent and suppressing the non-quantificational reading of arguments. In fact, the passivization of unaccusative verbs depends both on the atelic interpretation imposed by the construction and on the habitual and repetitive event structure that the verb reflects in the experiential world. Therefore, it is essentially a matter of interaction between the construction and the verb. Moreover, traditional grammars fail to provide a clear explanation for the passivization of intransitive verbs based solely on lexical projection. Previous research has also overlooked the constructional meaning of the impersonal passive. For this reason, the present study treats the impersonal passive as a construction within the framework of Goldberg’s Construction Grammar in order to address the following questions: What is the constructional meaning of the German impersonal passive? How can an intransitive verb be passivized into the impersonal passive? And why can non-volitional unaccusative verbs be passivized in this way?
2. Theoretical Approaches
Chapter 2 introduces Construction Grammar. Section 2.1 defines Construction Grammar, and Section 2.2 explains the interaction between verbs and constructions, thereby providing a basis for analyzing the German impersonal passive from the perspective of Construction Grammar.
2.1. Construction Grammar
Goldberg proposed Construction Grammar, which is grounded in Frame Theory and experience-based linguistics. Goldberg (1995: p. 5) defines a construction as follows:
C is a CONSTRUCTION iff_{def} C is a form-meaning pair
She argues that a construction is defined as a form meaning pair and is characterized by its unpredictability. This means that its meaning cannot be derived from its components or from other constructions. Constructions are considered fundamental units of language, whereby both phrases and morphemes can be regarded as constructions.
According to Goldberg’s definition, passive structures can become entrenched as constructions due to their high frequency. It is also worth noting that the meaning of passive constructions cannot be derived from their component parts. For example, the meaning of the impersonal passive is not simply the sum of its lexical components.
According to Construction Grammar, constructions are characterized by productivity. They are abstract, schematic event frameworks grounded in life experiences. New information and content can be inserted into these frameworks, giving rise to various types of expressions. This indicates that linguistic expressions or structures are formed in a top down manner, which contradicts the notion of lexical projection in generative grammar. It should also be noted that Construction Grammar is generative but not transformational. It dispenses with underlying representations and is monostratal, similar to Cognitive Grammar (Langacker, 1991, 2008). For example, passive constructions are not derived from their corresponding active constructions, which stands in contrast to the accounts offered by generative grammar. Instead, passive constructions are schematized based on life experiences.
Construction Grammar is based on a semantic approach that emphasizes speaker-centered constructions of situations and does not assume a strict separation between the lexicon and syntax. Both lexical and syntactic constructions are regarded as declarative form-meaning pairs. Likewise, the boundary between semantics and pragmatics is eliminated, so that information about focus, topic, or register becomes part of the constructions (Goldberg, 1995: p. 6). According to Goldberg (1995: p. 3), the functions of constructions include discourse functions and communicative functions. It is precisely these pragmatic elements that contribute to the understanding of constructions. For example, the impersonal passive cannot be understood solely at the syntactic level without considering context and background information.
2.2. The Interaction between Verb and Construction
In the interaction between the construction and the verb, each argument role of the construction fuses with the corresponding participant role of the verb, thereby forming concrete expressions. Fusion here serves to capture the semantic constraints on the participant roles associated with the verb and the argument roles of the construction, as opposed to denoting the fusion of gaps within a single lexical entry (Goldberg, 1995: p. 50).
In the fusion of participant roles and argument roles, the principle of semantic consistency must be observed. For example, the constructional meaning of resultative constructions is that the agent causes the resultant state of the patient. In this case, the abstract argument structure carries the semantic roles of agent, patient, and goal, which fuse with the participant roles of the verb. The corresponding syntactic forms are subject, direct object, indirect object, and oblique. Together, the semantic roles and the syntactic slots form the construction, in which form and meaning correspond to each other.
However, argument structure can alter the semantic roles of the verb. When constructional meaning and lexical meaning come into conflict, constructional coercion enriches the lexeme by removing conflicting components of the word meaning through pruning or shading, or by adding compatible semantic elements, so as to adapt it to the constructional meaning (cf. Goldberg, 1995: p. 57; Michaelis, 2003, 2004, 2005). For example, the verb “sneeze” is monadic and has only the agent role of “sneezer”. However, in the interaction between the caused-motion construction and the verb “sneeze”, the semantic roles of patient and goal are added, giving rise to the sentence: Sam sneezed the napkin off the table. In this sentence, “the napkin” functions syntactically as the object and semantically as the patient, while the prepositional phrase “off the table” is syntactically an oblique, with “the table” serving semantically as the goal.
As with nouns, verbs also lexically determine which aspects of their frame-semantic knowledge are obligatorily profiled (Goldberg, 1995: p. 44). This indicates that the profiling of certain aspects of a verb’s frame-semantic knowledge facilitates the interaction between construction and verb. Specifically, the verb’s semantic frame expands its range of usage, and it is the lexical meanings grounded in world knowledge that first make this interaction possible. In the example above, it is profiled that the verb “sneeze” produces an airflow that represents a dynamic force. Consequently, the monadic verb can be embedded into the caused-motion construction.
In the interaction between verb and construction, a specific participant role can be shaded. The passive construction serves to defocus the highest-ranking participant role of a verb, so that this role can at most be realized as an adjunct (Goldberg, 1995: p. 57). This presupposes a thematic hierarchy that determines the ranking of the roles:
Agent, Causer > Recipient, Experiencer > Instrument > Patient, Theme > Location, Source, Goal
The roles listed in the thematic hierarchy are general argument roles (Dowty, 1986), which represent superordinate types relative to the participant roles of the verb. Since participant roles are typically instantiations of such general roles, the hierarchy establishes a partial order that does not relate all roles to one another. Consequently, the passive applies only to verbs that have at least two roles, with one role ranked higher in the hierarchy than the other.
3. The Passivization of Intransitive Verbs in the German Impersonal Passive
Chapter 3 first examines how the verb, particularly the unaccusative verb, is passivized through its interaction with the passive construction. Accordingly, the conditions governing the passivization of intransitive verbs are analyzed.
Interaction between the Verb and the Impersonal Passive
According to Goldberg (1995: p. 15), the impersonal passive is a construction whose meaning cannot be predicted from its component parts. She criticizes the claim that the passivization of unaccusative verbs is justified by lexical ambiguity (i.e., an atelic and a telic reading), given that auxiliary selection with “sein” already indicates the telicity of unaccusative verbs at the lexical level:
A theory that posited two lexical items with opposite Aktionsart specifications would not be able to predict these facts about auxiliary selection. (Goldberg, 1995: p. 15)
Against the hypothesis that unaccusative verbs exhibit two readings (an atelic and a telic one), Beatrice Primus (2011) employs the event structural repair, a mechanism that has already been attested in psychological studies (Todorova et al., 2000; Pylkkänen & McElree, 2006). An originally telic unaccusative verb can cancel its telic meaning and thereby acquire an atelic reading, namely a repetitive or habitual interpretation:
(5) Das Mädchen ist drei Stunden lang gesprungen. (The girl jumped for three hours.)
She also provided evidence for the repair hypothesis in acceptability experiments. Empirically, this results in a clear drop in acceptability from non-volitional atelic verbs to non-volitional telic verbs in the impersonal passive, because unaccusative verbs can only be passivized through event structural repair. The assumption of lexical ambiguity for unaccusatives cannot properly account for their passivization.
She adds that the lexical aspect of an event is determined by the reference type of the relevant arguments in the event structure. Quantized referential arguments determine the telic semantics of an event, while non-quantized referential arguments determine its atelic semantics.
She further explains that through referential argument demotion, the suppressed argument becomes a non-quantized component. This leads to originally telic unaccusative verbs receiving a repetitive, habitual atelic interpretation. Therefore, she argues that the atelic semantics of the impersonal passive is attributable neither to the construction itself nor to lexical ambiguities of unaccusative verbs, but rather to a construction-independent felicity constraint, which requires a quantized expression in an event-structure-sensitive function.
However, her view is one-sided in attributing the atelic reading of unaccusative verbs to the downgrading of the agent, while neglecting the constructional meaning of the impersonal passive. This is because agent demotion is not a distinctive feature of the impersonal passive. It also occurs in the personal passive. In the personal passive, however, the downgrading of the agent does not necessarily cause the suppressed agent to take on a non-quantized character. Consequently, no causal relationship can be established between the downgrading of the agent and the non-quantized nature of the agent argument. In the personal passive, a quantized agent can be realized in the form of a prepositional phrase:
(6) Er wurde gemalt von Franz Radziwill und war 1995 ein Briefmarkenmotiv. (It was painted by Franz Radziwill and was the motif of a postage stamp in 1995.)
(7) Es ist auch schwer herauszufinden, ob das unterdrückte Agens quanteltet oder nicht quanteltet ist: (It is also difficult to determine whether the suppressed agent is quantized or non-quantized:)
(8) Diese wurde später durch eine Stahlkonstruktion verstärkt, und das Dach wurde repariert. (This was later reinforced by a steel structure, and the roof was repaired.)
It is noteworthy that in the following impersonal passive constructions, the adverbial “früh” (early) with telic semantics is compatible with the unaccusative verb “sterben” (to die). This cannot be adequately explained by the pragmatic mechanism of event structural repair proposed by Beatrice Primus, which is determined by non-quantized suppressed arguments and durative temporal adverbials. This shows that a purely lexical or pragmatic explanation of the atelic reading of the impersonal passive without considering constructional coercion is insufficient:
(9) Es wird geliebt, mehr noch gehasst, und es wird früh gestorben. (It is loved, even more so hated, and people die young.)
According to Arnett (2002), the past participle in the impersonal passive is formed through the integration of the verb stem with PERF. In her view, PERF here should denote all states within the process rather than only the final state, which gives the impersonal passive an atelic reading. However, this excludes the possibility that unaccusative verbs can be used in the impersonal passive, since they require the auxiliary “sein” and their past participle marks a punctual, telic single-participant event. This makes it clear that it is inappropriate to completely disregard the constructional meaning of the impersonal passive and to analyze the passivization of intransitive verbs in isolation.
The impersonal passive is grounded in human beings’ experience of habitual and atelic activities. It serves as a generalized description of an atelic, non-punctual event or state embedded within an abstract or concrete setting, with the focus placed on the processuality and holistic viewing of the action rather than on a concrete, bounded, punctual event. It must be clear that the emergence of the atelic semantics in the impersonal passive construction essentially relies on a metonymic mechanism. At the lexical level, the verb represents a one-time, bounded, telic event, whose understanding is based on sequential scanning of temporal sequences in mental processing. However, in order to transform this into a repeated, habitual result state, a metonymy of the type “ENTITY FOR ACTION” is required. That is, a concrete event entity stands for the abstract action pattern that constitutes it. In this case, mental processing also switches to summary scanning, which generalizes the entire situation.
The argument structure of the German impersonal passive construction is very simple: it contains only an agent role, and this agent is completely suppressed. Since the verb on which the construction is based is intransitive and does not itself carry a patient argument, there is no patient role in the construction. At the syntactic level, the agent cannot appear, and the subject position is typically occupied by the expletive subject es, but es does not bear any semantic role. In other words, the construction profiles neither the agent nor the patient; its semantic focus lies entirely on the action or process itself.
The passivization process of German intransitive verbs is essentially a fusion and interaction between the impersonal passive construction and the intransitive verb. Concretely, the argument structure of the impersonal passive must be integrated with the thematic structure of the intransitive verb. In this process, the activity participant in the construction fuses with the single agent role assigned by the verb, ultimately leading to an abstract and generalizing description of the activity while simultaneously suppressing the telicity and punctuality of the event.
For example, in the interaction between the verb arbeiten (to work) and the impersonal passive construction, passivization can easily occur. This is because the verb denotes a process without an inherent endpoint, which corresponds precisely to the constructional meaning of the impersonal passive―namely, the generalization of a habitual and atelic activity. At the same time, the agent role is suppressed, as the construction shifts the focus away from the initiator of the action and onto the process itself.
The constructional meaning of the impersonal passive foregrounds habitual and atelic activities. According to the semantic coherence principle, this constructional meaning should be highly congruent with the atelic semantics of the verbs. To test this hypothesis, this study randomly collected 135 natural sentences containing the impersonal passive, covering both written and spoken registers and involving various text types such as narrative, descriptive, and expository. From the 135 example sentences, all verbs appearing in the impersonal passive were extracted, yielding a total of 368 verb tokens and 136 distinct verb types.
Based on whether an event has an inherent natural endpoint as the core semantic feature, this paper classifies verbs into two major categories. The first category consists of verbs that describe activities, processes, or states, where the event lacks an inherent natural endpoint and is characterized by the semantic features [?telic] and [+durative], such as “tanzen” (to dance), “schlafen” (to sleep), and “arbeiten” (to work). The second category comprises telic verbs, which describe achievements or accomplishments, where the event has a clear inherent endpoint, as exemplified by “sterben” (to die) and “blühen” (to bloom).
Among the total 368 verb tokens, atelic verbs occurred 330 times, accounting for 89.7% of the total, while telic verbs occurred 38 times, accounting for 10.3% of the total, as shown in the following Table 1:
![]()
Table 1. Distribution of verb tokens in the impersonal passive.
As shown in Table 2, among the 136 distinct verb types, 112 verbs (82.4%) are atelic in semantic nature, while 24 verbs (17.6%) are telic: “sterben” (to die) and “blühen” (to bloom):
![]()
Table 2. Distribution of atelic and telic verb types in the impersonal passive.
From the total of 368 verb tokens, atelic verbs with a frequency of two or more occurrences were selected, yielding a total of 17 verbs:
![]()
Figure 1. Frequency of selected atelic verbs in the impersonal passive.
The above Figure 1 shows the frequencies of selected atelic verbs in the impersonal passive. The most frequent verb is “singen” (8 occurrences), followed by “lachen” (5) and “trinken” (4). Notably, “schwitzen” (to sweat) appears twice, representing a physiological process that is inherently atelic, as sweating lacks a natural endpoint and can continue indefinitely. All listed verbs denote activities, processes, or states without an inherent endpoint, aligning with the atelic semantics of the impersonal passive construction.
Telic verbs generally occur with low frequency in the corpus, mostly appearing only once. Only one telic verb, “sterben” (to die), occurs three times, and all its occurrences appear in habitual or generic contexts. Other telic verbs, such as “blühen” (to bloom), “sprengen” (to blast), and “fallen” (to fall), occur only once. Although these verbs are inherently telic, once integrated into the impersonal passive construction, their telic semantics are suppressed, and the focus shifts to an abstracted description of the activity.
In the first example below, although “sprengen” (to blast) conveys telic semantics, the sentence as a whole presents an objective description of a natural phenomenon. Thus, the focus is no longer on a specific blasting event, but rather on the habitual activity of blasting the blocked ice. In the second example below, which describes the grand historical context of post-war reconstruction, the verb does not refer to a specific act of dismantling, but rather to a general description of the habitual and widespread dismantling activities that took place during the post-war reconstruction process:
(10) An der Loreley staute sich das Eis so stark, dass keine Eisbrecher mehr durchkamen; es wurde gesprengt. (At the Loreley, the ice accumulated so heavily that no icebreakers could get through anymore; so blasting was carried out.)
(11) Es wurde demontiert und man begann erst im Jahr 1951 wieder mit der Produktion. (Dismantling was carried out, and production was not resumed until 1951.)
First, all high-frequency verbs are atelic, with “singen” (sing), “tanzen” (dance), and “lachen” (laugh) occurring most frequently. The activities denoted by these verbs share the semantic features [+durative] and [-resultative], which naturally align with the constructional meaning of the impersonal passive that foregrounds habitual and atelic activities. Second, even when telic verbs do occur, their frequency is significantly lower than that of atelic verbs. The only telic verb that appears three times, “sterben” (die), is accepted only in habitual or generic contexts (e.g., “people die young”), which conversely demonstrates the construction’s preference for atelic semantics: even when a telic verb enters the construction, it must be forced into a habitual interpretation. Third, the distribution of high-frequency verbs further supports the core conclusion of this study: in real language use, the impersonal passive systematically selects atelic verbs, and a stable semantic coherence exists between the two.
4. The Conditions for Passivization in German
In Section 4.1, the hypothesis is put forward that agentivity is the condition for passivization. In Section 4.2, this hypothesis is tested against the corpus data.
4.1. Hypothesis: Agentivity Is the Condition for Passivization
There is currently no consensus regarding the semantic features of the impersonal passive. On the one hand, some scholars argue that the impersonal passive is primarily characterized by volitionality (Kirsner, 1979, Perlmutter, 1978, Drosdowski, 1984: p. 183, Moorcroft, 1985, Fagan, 1992, Shannon, 1992, Arnett, 2002). Volitionality is one of the parameters of transitivity established by Hopper and Thompson (1980). For this reason, the impersonal passive exhibits features of transitivity. To demonstrate the volitionality of the impersonal passive, Arnett cites a typical example from Curme (1905: p. 338):
(12) Für den lieben König und Herren wird alles getan, wird treulich gekämpft, wird willig geblutet, wird freudig in den Tod gegangen, für ihn wird mehr als gestorben. (For the dear king and lord, everything is done, loyally fought, willingly bled, joyfully gone to death; for him, more than death is endured.)
The above example shows that an intransitive verb cannot be passivized if a volitional context has not been established in the sentence (Fagan, 1992, Shannon, 1992, Arnett, 2002). The verbs mentioned above are all intransitive.
The above example indicates that volitionality is a factor in the generation of the impersonal passive from an intransitive verb. Each action denoted by the verb is modified by an adverb. For example, “kämpfen” (to fight), “bluten” (to bleed), “in den Tod gehen” (to go to death), and “sterben” (to die) are each modified by “treulich” (loyally), “willig” (willingly), “freudig” (joyfully), and “mehr als” (more than), respectively. This indicates how each action is performed.
In reality, however, non-volitional verbs can also occur in the impersonal passive. According to Dowty (1991: p. 609), it is often used to achieve fanciful, metaphorical, humorous, or other non-literal effects.
Contrary to Dowty’s view (Dowty, 1991), the examples below are neither metaphorical nor humorous:
(13) Oft bleiben da Touristen auch länger und verheizen das ganze Holz, sodass im Ernstfall dann gefroren wird. (Tourists often stay there longer and burn up all the wood, so that in an emergency, people end up freezing).
(14) Besondere Sensibelchen aus der Duftabteilung rümpfen immer die Nase. Weil im Sommer so viel gestunken wird. (The particularly sensitive ones in the fragrance department always turn up their noses. Because there is so much stench in the summer.)
(15) Aber geblüht wird nur, wenn die Pflanze auch etwas älter ist. Es daurt einige Jahre, bis die Blüten kommen. (But flowering only occurs when the plant is a bit older. It takes several years for the flowers to come.)
To explain the above examples, Beatrice Primus (2011: p. 301) redefined the features of agentivity: the suppressed argument in the impersonal passive must exhibit at least one of the following agentivity features: a control feature, sentience, or self-induced movement, as specified further below:
This agentivity constraint presupposes a multidimensional concept of agent. According to Dowty, it comprises volitionality, causation, sentience (i.e., perception, sensation, emotion, or other mental states), and movement.
With regard to control (volitionality or intentionality), Tomasello (2005) has outlined its dimensions as follows:
1) the event-specific goal-directedness of a system involved in an event, whereby the bringing about of the event or its omission also counts as a goal;
2) the choice of means to achieve this goal, or an action plan;
3) the perceptual monitoring of the action goal.
In the present study, we align with Primus’s view that agentivity is multidimensional, encompassing control, sentience, and self-induced movement.
The feature of control refers to volitionality and intentionality, that is, the conscious initiation of an event. Sentience pertains to perception: the external world stimulates a cognitive entity, which responds accordingly. Although perception does not necessarily involve conscious or volitional control, the resulting action is dynamic, and it is precisely this dynamism that characterizes agentivity.
Self-induced movement refers to how movements are carried out automatically and in accordance with human-designed plans. For example, an automatic device operates according to predefined programs, which are ultimately shaped by human will and intentionality.
4.2. Empirical Analysis Based on Corpus Data
To test the hypothesis that agency is a necessary condition for the formation of the impersonal passive, a controlled corpus analysis was conducted. The data were extracted exclusively from the DWDS (Digitales Wörterbuch der deutschen Sprache) corpus. The search was performed using the DWDS online interface, with the query pattern “es wurde” + past participle applied to the core corpus (“Kernkorpus”), which contains a balanced selection of 20th- and 21st-century German texts, including newspapers, fiction, academic prose, and transcribed spoken language. A total of 500 raw hits were retrieved from the period 2000-2020. After manual filtering to exclude false positives (e.g., “Personal Passive”), 138 unique verb types forming genuine impersonal passives with a covert, generic argument remained. Each verb was independently coded by two annotators for agentivity features (control, sentience, self-induced motion) based on Primus (2011). Inter-annotator agreement reached κ = 0.94, indicating high reliability.
The present study builds on the theoretical framework of agentivity proposed by Beatrice Primus (2011), whose core features are control, sentience, or self-induced motion. On this basis, agentivity is further subdivided into strong and weak manifestations. Using 138 authentic corpus examples, this distinction is empirically investigated with the aim of developing a continuum model of agentivity that accounts for the formation of the impersonal passive in German. The following Table 3 presents the distribution of verbs with strong and weak agentivity:
![]()
Table 3. The distribution of verbs with “strong” and “weak” agentivity in the impersonal passive.
Strong agentivity is intentional, volitional, and goal-directed. Concrete examples are presented below:
(16) Es wird gebastelt, gesungen und vorgelesen und bereits früh werden sprachliche Fähigkeiten geschult. (Crafting, singing, and reading aloud take place, and linguistic skills are trained from an early age.)
(17) Es wird Musik gemacht, es wird getanzt. (Music is being made, dancing is taking place.)
(18) Es gab keine festen Preise, es wurde gehandelt. (There were no fixed prices; bargaining took place.)
The examples above illustrate how atelic activities are carried out intentionally, volitionally, and in a goal-directed manner. The suppressed agent is understood to be a conscious human being. In these cases, the agent role of the verb merges with that of the impersonal passive construction. Even though the agent is not overtly expressed, its agentivity can still be inferred from the verb and the construction itself.
Weak agentivity is classified into the following three types:
1) Sentience;
2) Biological programs governing the sequence of events;
3) Physical and mechanical programs.
With regard to sentience, the term refers to the perceptions of living beings. Here, goal-directed biological programs are at work in bodily reactions. The stimulus excites the body, and in doing so, the human perceives it. For this reason, it concerns an interaction between the body and the environment that leads to the human perceiving the stimulus. The specific verbs are presented below:
frieren (freeze), atmen (breathe), bibbern (shiver), schwitzen (sweat), keuchen (pant), röcheln (wheeze), schnarchen (snore), gähnen (yawn), niesen (sneeze), husten (cough), sterben (die), bluten (bleed), quietschen (squeak), stinken (stink)
For example, the verb “schwitzen” (to sweat) involves a physiological process in which a sentient human being unconsciously reacts physically to a stimulus. Here, the suppressed argument can be regarded as an experiencer. It is noteworthy that the bodily reactions are generalized in the impersonal passive. From this, the habitual and atelic reading arises.
It is also worth noting that simple living beings can biologically program the sequence of events. For example, the verb “blühen” (to bloom) indicates how flowering plant species biologically program the blooming event in a targeted manner. Here, the biological programs are generalized in the impersonal passive, which leads to a habitual and atelic semantics.
The action of an automaton such as a sprinkler or a switch satisfies the criteria of goal-directedness and perceptual monitoring as defined by Beatrice Primus. For the automatic bringing about and omission of an event, automatons are consciously and purposefully manufactured by humans who possess a rudimentary, event-related perceptual system that triggers an event and controls its execution when stimulated by environmental changes:
einstellen (adjust), umschalten (switch)
In these cases, the suppressed argument is not a human but a manufactured control system. The impersonal passive generalizes the automatic, programmatic nature of the event. The presence of design-directedness (the automaton was built to perform this function) provides just enough agentivity to license the construction.
The distribution in Table 3 suggests not a binary distinction but a continuum of agentivity. At the strong end, verbs of conscious, intentional action routinely appear in the impersonal passive. At the weak end, verbs of sentience and biological/mechanical programming appear but are less frequent (14.5%). Beyond the weak end, verbs of pure change of state or existence do not appear at all. This gradient pattern supports an inclusion threshold hypothesis: only verbs whose subject argument possesses at least one of Primus’s features (control, sentience, or self-induced motion) can have that argument suppressed in the impersonal passive. Verbs lacking all three features are excluded.
5. Summary and Outlook
In summary, the constructional meaning of the impersonal passive is grounded in a specific conceptualization of the world, and it is this conceptual basis that determines the suppression of the agent. By backgrounding the agent, the construction directs attention toward a generalized description of habitual and repeated activities. Unaccusative verbs, by their very nature, are capable of expressing habitual and repetitive actions. It is precisely this inherent semantic feature that fuses with the constructional meaning of the impersonal passive, and it is through this fusion that the passivization of unergative verbs is ultimately achieved. This fusion is not a matter of lexical ambiguity, but rather a genuine interaction between the verb’s inherent aspectual properties and the construction’s generalized, atelic semantics. Following the empirical verification conducted in this study, agentivity is confirmed as a necessary condition for the passivization of intransitive verbs in German. Importantly, agentivity is not a binary property but exists along a continuum ranging from “weak” to “strong”. This continuum encompasses a broad spectrum of agent-related features, including volition and intentionality at the strong end, sentience and perceptual responsiveness in the middle, and, at the weaker end, biological programs that govern species-specific event sequences (such as flowering in plants), as well as physical or mechanical programs that characterize the operation of automatons and human-designed systems.
Future research aims to test the agentivity continuum hypothesis from a broader cross-linguistic perspective while employing psycholinguistic methods to investigate the cognitive processing involved in the passivization of unaccusative verbs, including the metonymic mechanism of “entity for action” and the corresponding shift in cognitive scanning modes. In addition, corpus data should be enriched by incorporating colloquial and diachronic sources, which will help reveal the distribution and semantic properties of the impersonal passive in different registers and provide deeper insight into the dynamic interaction between verbs and constructions.